Comprehensive Analysis of Machine Learning Solutions for Energy Demand Forecasting in Distributed Systems

In the rapidly-evolving context of energylandscape, and in the transition to a carbon-neutraleconomy and deployment of plug-in electricvehicles (PEVs), new challenges are being placedon small power systems. The power generation andconsumption are decentralized compared toconventional stable monolithic power plants, suchas we have many smaller, autonomous power unitsthat need to adapt to evolving energy demands. Theunpredictability of renewable energy combinedwith fluctuations in energy needs makesconventional grid management techniques andforecasting measures, for instance ARIMA andExponential Smoothing, increasingly inapplicable.These conventional approaches, while useful insimpler, stable grid environments, struggle to adaptto complex, non-linear dynamics of today's grids.